What are the Data of Philosophical Thought Experiments?

I argue that the data collected by philosophical thought experiments--the “intuitions” we appeal to regarding hypothetical situations constructed to test philosophical hypotheses--should be construed as metalinguistic, about what we are inclined to say. For one thing, although concepts cannot be directly perceived, they are reflected in our use of the words associated with them; moreover, our
beliefs about what we would say in a given situation are typically less vulnerable to objection than our object-level beliefs about the situation itself. I submit that these metaphilosophical considerations might ultimately provide evidence against semantic contextualism.